Files
rengine/shared/models/endpoint.py
T
Yogesh OjhaandClaude Opus 5 0b28f430af fix(endpoints): a long path segment no longer fails URL discovery
`normalize_path` caps a path at MAX_PATH_LENGTH, but `split_path` takes the last
segment as the filename and that column is varchar(300) — a 1,400-character segment
with no slash in it failed the insert, poisoned the transaction and cost the whole
stage. gov.np hit it once the HTTP probe stopped truncating itself and handed URL
discovery 15,098 web assets instead of 2,687.

MAX_HOST_LENGTH and MAX_FILENAME_LENGTH join the caps already in
shared/definitions/endpoints.py, `split_path` applies the filename cap and
`parse_url` rejects an over-long host rather than storing a truncated one, and the
model's max_length values now read from those constants instead of repeating the
numbers. The batch insert runs in a savepoint and retries row by row, so a row
postgres refuses costs that row and lands in UpsertResult.rejected — the same rule
the HTTP probe now follows.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DPueZXQMbjLXMokXeA8LFP
2026-09-06 06:22:48 +05:30

461 lines
14 KiB
Python

import uuid
from datetime import datetime
from pydantic import BaseModel
from pydantic import Field as PydanticField
from sqlalchemy import Column, Text
from sqlalchemy.types import JSON
from sqlmodel import Field, SQLModel, UniqueConstraint
from shared.definitions.asset_query import MAX_QUERY_LENGTH
from shared.definitions.endpoints import (
MAX_FILENAME_LENGTH,
MAX_HOST_LENGTH,
MAX_PATH_LENGTH,
MAX_URL_LENGTH,
EndpointClass,
EndpointSource,
)
from shared.definitions.vulnerabilities import CoverageStatus
from shared.models.asset_query import MatchEvidence, QueryError
from shared.utils.datetime import utc_now
def _json_list() -> Field:
return Field(default_factory=list, sa_column=Column(JSON, nullable=False))
def _json_dict() -> Field:
return Field(default_factory=dict, sa_column=Column(JSON, nullable=False))
def _text() -> Field:
return Field(default=None, sa_column=Column(Text, nullable=True))
class Endpoint(SQLModel, table=True):
__tablename__ = "endpoints"
__table_args__ = (
UniqueConstraint("scan_id", "signature", name="uq_endpoint_scan_signature"),
)
id: uuid.UUID = Field(default_factory=uuid.uuid4, primary_key=True, index=True)
scan_id: uuid.UUID = Field(foreign_key="scans.id", index=True, ondelete="CASCADE")
target_id: uuid.UUID = Field(
foreign_key="targets.id", index=True, ondelete="CASCADE"
)
project_id: uuid.UUID = Field(foreign_key="projects.id", index=True)
# structural identity: path shape plus parameter names, never parameter values
signature: str = Field(max_length=64, index=True)
# where it is
url: str = Field(max_length=MAX_URL_LENGTH)
host: str = Field(max_length=MAX_HOST_LENGTH, index=True)
port: int = Field(default=443)
scheme: str = Field(default="https", max_length=8)
path: str = Field(max_length=MAX_PATH_LENGTH)
dir_path: str = Field(max_length=MAX_PATH_LENGTH, index=True)
filename: str | None = Field(default=None, max_length=MAX_FILENAME_LENGTH)
extension: str | None = Field(default=None, max_length=10, index=True)
depth: int = Field(default=0, index=True)
# what it accepts
params: list = _json_list()
param_count: int = Field(default=0, index=True)
param_samples: list = _json_list()
variants: int = Field(default=1)
more_variants: bool = Field(default=False)
methods: list = _json_list()
# who says so
sources: list = _json_list()
primary_source: str = Field(
default=EndpointSource.OTHER.value, max_length=24, index=True
)
discovery: dict = _json_dict()
found_on: str | None = Field(default=None, max_length=2000)
# what it answered, if anything asked
is_probed: bool = Field(default=False, index=True)
status_code: int | None = Field(default=None, index=True)
content_type: str | None = Field(default=None, max_length=255)
content_length: int | None = Field(default=None)
title: str | None = Field(default=None, max_length=1000)
words: int | None = Field(default=None)
lines: int | None = Field(default=None)
response_time: float | None = Field(default=None)
redirect_location: str | None = Field(default=None, max_length=2000)
content_hash: str | None = Field(default=None, max_length=80, index=True)
tech: list = _json_list()
# how it reads
endpoint_class: str = Field(
default=EndpointClass.OTHER.value, max_length=16, index=True
)
interest: list = _json_list()
# correlation ids, resolved at write time; no FK so a row outlives its asset
http_asset_id: uuid.UUID | None = Field(default=None, index=True)
subdomain_id: uuid.UUID | None = Field(default=None, index=True)
archive_last_seen: datetime | None = Field(default=None)
discovered_at: datetime = Field(default_factory=utc_now, index=True)
created_at: datetime = Field(default_factory=utc_now)
class EndpointCoverage(SQLModel, table=True):
"""One provider's account of what it covered. A null count means unknown, never zero."""
__tablename__ = "endpoint_coverage"
id: uuid.UUID = Field(default_factory=uuid.uuid4, primary_key=True, index=True)
scan_id: uuid.UUID = Field(foreign_key="scans.id", index=True, ondelete="CASCADE")
target_id: uuid.UUID = Field(
foreign_key="targets.id", index=True, ondelete="CASCADE"
)
project_id: uuid.UUID = Field(foreign_key="projects.id", index=True)
source: str = Field(max_length=24, index=True)
tool: str | None = Field(default=None, max_length=40)
status: str = Field(default=CoverageStatus.COMPLETED.value, max_length=16)
hosts_total: int = Field(default=0)
hosts_scanned: int | None = Field(default=None)
hosts_dropped: list = _json_list()
urls_found: int | None = Field(default=None)
urls_stored: int | None = Field(default=None)
urls_probed: int | None = Field(default=None)
pages_fetched: int | None = Field(default=None)
depth_reached: int | None = Field(default=None)
errors: int | None = Field(default=None)
capped: bool = Field(default=False)
cap_reason: str | None = Field(default=None, max_length=200)
command: str | None = _text()
error: str | None = Field(default=None, max_length=2000)
started_at: datetime = Field(default_factory=utc_now)
ended_at: datetime | None = Field(default=None)
duration_seconds: float | None = Field(default=None)
class SourceEvidence(BaseModel):
"""Why one provider believes this endpoint exists."""
source: str
label: str
kind: str
detail: str | None = None
found_on: str | None = None
observed_at: datetime | None = None
class EndpointRead(BaseModel):
id: uuid.UUID
scan_id: uuid.UUID
target_id: uuid.UUID
signature: str
url: str
host: str
port: int
scheme: str
path: str
dir_path: str
filename: str | None = None
extension: str | None = None
depth: int
params: list[str] = PydanticField(default_factory=list)
param_count: int = 0
variants: int = 1
more_variants: bool = False
methods: list[str] = PydanticField(default_factory=list)
sources: list[str] = PydanticField(default_factory=list)
primary_source: str
evidence: list[SourceEvidence] = PydanticField(default_factory=list)
found_on: str | None = None
is_probed: bool = False
status_code: int | None = None
content_type: str | None = None
content_length: int | None = None
title: str | None = None
words: int | None = None
lines: int | None = None
response_time: float | None = None
redirect_location: str | None = None
tech: list[str] = PydanticField(default_factory=list)
endpoint_class: str
interest: list[str] = PydanticField(default_factory=list)
http_asset_id: uuid.UUID | None = None
subdomain_id: uuid.UUID | None = None
archive_last_seen: datetime | None = None
discovered_at: datetime
is_new: bool = False
matched_in: list[MatchEvidence] = PydanticField(default_factory=list)
class EndpointDetail(EndpointRead):
param_samples: list = PydanticField(default_factory=list)
discovery: dict = PydanticField(default_factory=dict)
content_hash: str | None = None
siblings: int = 0
class EndpointFilter(BaseModel):
q: str | None = PydanticField(default=None, max_length=MAX_QUERY_LENGTH)
host: str | None = None
dir_path: str | None = None
subtree: bool = True
endpoint_class: str | None = None
source: str | None = None
interest: str | None = None
status_class: str | None = None
probed: bool | None = None
new: bool = False
hide_static: bool = False
sort: str = "path"
direction: str = "asc"
page: int = 1
size: int = 50
def has_facets(self) -> bool:
return any(
(
self.host,
self.dir_path,
self.endpoint_class,
self.source,
self.interest,
self.status_class,
self.probed is not None,
self.new,
self.hide_static,
)
)
class EndpointPage(BaseModel):
items: list[EndpointRead] = PydanticField(default_factory=list)
total: int = 0
total_capped: bool = False
page: int = 1
size: int = 50
error: QueryError | None = None
class EndpointFacet(BaseModel):
value: str
label: str
count: int
class EndpointFacets(BaseModel):
endpoint_class: list[EndpointFacet] = PydanticField(default_factory=list)
source: list[EndpointFacet] = PydanticField(default_factory=list)
interest: list[EndpointFacet] = PydanticField(default_factory=list)
status_class: list[EndpointFacet] = PydanticField(default_factory=list)
extension: list[EndpointFacet] = PydanticField(default_factory=list)
host: list[EndpointFacet] = PydanticField(default_factory=list)
total: int = 0
static_total: int = 0
class TreeLeaf(BaseModel):
"""The one endpoint a folder holds when that folder is only its own index."""
id: uuid.UUID
url: str
host: str
path: str
params: list[str] = PydanticField(default_factory=list)
param_count: int = 0
endpoint_class: str
is_probed: bool = False
status_code: int | None = None
content_length: int | None = None
sources: list[str] = PydanticField(default_factory=list)
interest: list[str] = PydanticField(default_factory=list)
class TreeNode(BaseModel):
"""One directory or host in the site tree, counted through the same filter as the table."""
key: str
name: str
path: str
host: str | None = None
kind: str = "directory"
depth: int = 0
direct_count: int = 0
subtree_count: int = 0
child_count: int = 0
hosts: int = 1
status_mix: dict[str, int] = PydanticField(default_factory=dict)
class_mix: dict[str, int] = PydanticField(default_factory=dict)
sources: list[str] = PydanticField(default_factory=list)
interest: list[str] = PydanticField(default_factory=list)
has_params: bool = False
params: int = 0
verified: int = 0
unprobed: int = 0
new_count: int = 0
gone_count: int = 0
anomaly: str | None = None
archive_only: bool = False
glyph: str = "folder"
sample_url: str | None = None
leaf: TreeLeaf | None = None
query: str
children: list["TreeNode"] = PydanticField(default_factory=list)
lazy: bool = False
folders: int = 0
top_folders: list[str] = PydanticField(default_factory=list)
class HostPage(BaseModel):
"""One page of hosts with their rollups; a host's folders load when it is opened."""
items: list[TreeNode] = PydanticField(default_factory=list)
total: int = 0
total_endpoints: int = 0
page: int = 1
size: int = 50
error: QueryError | None = None
class MergedLeaf(BaseModel):
"""One path inside a folder, folded across every host that serves it."""
key: str
path: str
name: str
params: list[str] = PydanticField(default_factory=list)
param_count: int = 0
endpoint_class: str
hosts: int = 0
endpoints: int = 0
status_mix: dict[str, int] = PydanticField(default_factory=dict)
unprobed: int = 0
interest: list[str] = PydanticField(default_factory=list)
sources: list[str] = PydanticField(default_factory=list)
host_names: list[str] = PydanticField(default_factory=list)
new_count: int = 0
sample_id: uuid.UUID
sample_url: str
sample_status: int | None = None
query: str
class MergedLeafPage(BaseModel):
items: list[MergedLeaf] = PydanticField(default_factory=list)
total: int = 0
truncated: bool = False
class EndpointTree(BaseModel):
mode: str = "host"
nodes: list[TreeNode] = PydanticField(default_factory=list)
total_endpoints: int = 0
total_nodes: int = 0
truncated: bool = False
error: QueryError | None = None
class CoverageRead(BaseModel):
id: uuid.UUID
source: str
label: str
tool: str | None = None
status: str
hosts_total: int = 0
hosts_scanned: int | None = None
hosts_dropped: list[str] = PydanticField(default_factory=list)
urls_found: int | None = None
urls_stored: int | None = None
urls_probed: int | None = None
pages_fetched: int | None = None
depth_reached: int | None = None
errors: int | None = None
capped: bool = False
cap_reason: str | None = None
error: str | None = None
started_at: datetime
ended_at: datetime | None = None
duration_seconds: float | None = None
class EndpointSummary(BaseModel):
total: int = 0
probed: int = 0
live: int = 0
with_params: int = 0
interesting: int = 0
hosts: int = 0
new: int = 0
gone: int = 0
previous_scan_id: uuid.UUID | None = None
previous_scan_at: datetime | None = None
by_class: dict[str, int] = PydanticField(default_factory=dict)
by_source: dict[str, int] = PydanticField(default_factory=dict)
class GonePage(EndpointPage):
"""Endpoints the previous scan of this target recorded and this scan did not."""
previous_scan_id: uuid.UUID | None = None
previous_scan_at: datetime | None = None
class StructureFinding(BaseModel):
"""Something about the shape of the surface that is worth acting on."""
kind: str
label: str
detail: str
count: int
query: str
samples: list[str] = PydanticField(default_factory=list)
class PathSpread(BaseModel):
path: str
hosts: int
endpoints: int
query: str
class StructureLine(BaseModel):
key: str
label: str
detail: str | None = None
count: int
hosts: int = 0
query: str
class ScanStructure(BaseModel):
endpoints: int = 0
hosts: int = 0
probed: int = 0
directories: int = 0
max_depth: int = 0
with_params: int = 0
headline: str | None = None
findings: list[StructureFinding] = PydanticField(default_factory=list)
shared_paths: list[PathSpread] = PydanticField(default_factory=list)
interest: list[StructureLine] = PydanticField(default_factory=list)
by_class: list[StructureLine] = PydanticField(default_factory=list)
by_source: list[StructureLine] = PydanticField(default_factory=list)
class VerifyBranchRequest(BaseModel):
host: str = PydanticField(max_length=500)
dir_path: str | None = PydanticField(default=None, max_length=1500)
limit: int = PydanticField(default=500, ge=1, le=2000)
class VerifyBranchResponse(BaseModel):
queued: int = 0
unverified: int = 0
accepted: bool = False